Assessing the Demand for AI Integration in Public Administration : Qualitative Evidence and Policy Implications from the National Police Agency

Despite the rapid quantitative growth of Artificial Intelligence (AI) in the public sector, a significant gap remains between technological investment and field-level satisfaction. This study proposes a shift toward a 'Demand-based Paradigm' to overcome the limitations of conventional supply-centric technology adoption. Focusing on the Korean National Police Agency (KNPA), which has seen accelerated AI investment following the 2025 national strategy, this research aims to derive practical AI business models that reflect the actual needs of field officers. For this purpose, three levels of qualitative data were collected and analyzed. First, open-ended survey responses from 407 police officers were analyzed using WordCloud to identify baseline demands. Second, 124 proposals from the 'Science and Policing Idea Contest' (2021-2024) were categorized through LDA-based topic modeling. Third, the identified demands were refined into policy models through in-depth interviews with field advisory groups and experts. The findings identify several demand-based AI projects across four core domains: Investigation Support, Field Response Enhancement, Administrative Innovation, and Organizational Infrastructure. Key projects include AI for traffic accident fault analysis, real-time guidance on the use of force, on-premise report drafting assistants, and AI guardrails for ethical governance. This study criticizes the existing approach focused solely on technical feasibility and argues that a user-centric approach—prioritizing the work context and 'pain points' of practitioners—is the essential mechanism for ensuring budget efficiency and improving the quality of public safety services.